Chronic pain affects nearly 20% of the Canadian population and is now recognized as a disease by the World Health Organization. Moderate to severe pain is currently treated with opioids which are known to cause various side effects (constipation, nausea, respiratory depression, drowsiness) and strong dependence. A promising alternative is neurotensin, an endogenous neuromodulator peptide which induces analgesia by interacting with its two G-protein coupled receptors NTS1 and NTS2. However, NTS1 activation is also associated with hypotension and hypothermia. Selectivity towards NTS2 is, therefore, crucial. This was achieved in collaboration with Valence Discovery by exploiting artificial intelligence (AI). 923 patents and 1400 articles have been used to extract chemical structures and biological data of known NTS1 and NTS2 ligands. This data was then used to develop various machine learning models designed to screen a commercial bank of 8 million compounds (MolPort library), predicting affinity and selectivity. These multiparameter algorithms have allowed us to identify 61 potential hits. Competitive binding assays with 125I-labeled [Tyr3]-neurotensin were performed on membrane homogenates from cells stably expressing NTS1 or NTS2 to determine Ki values and receptor selectivity. 13 compounds were identified as active hits on NTS2, and none were able to bind to NTS1 (5 compounds exhibiting Ki affinities below 10 µM). By expanding the molecular diversity within the hit-derived class, we anticipate that our future SAR study will provide potent NTS2-selective small molecules suitable for in vivo pharmacological testing in pain models.